fix: keep tag-like text and numeric citations in assistant answers

[caitlon]

Problem

Assistant answers lost text that looks like a tag: x<y and y>z rendered as xz, because raw HTML was skipped. A line such as [1]: https://... in an answer became a markdown link definition, so every [1] citation lost its brackets. The local-source hint still called the corpus private.

Change

• Raw HTML in an answer is shown as text. The element allowlist and unwrapDisallowed stay, and rehype-raw is still absent, so no HTML element can be created from an answer.
• A numeric definition line ([n]: ..., also behind list or quote markers) has its opening bracket escaped before parsing, so it stays plain text and the citations keep working. Top-level code blocks and non-numeric definitions are untouched.
• New local-source hint in en and cs; the JSDoc of AssistantMarkdown describes how [n] actually renders.
• react-markdown listed in the frontend tech stack.

Verification

• npm run lint, npm run typecheck, npm run test:coverage (thresholds held).
• New tests for tag-like text and for definition lines in five layouts; each failed before the change. Raw <script>/<img onerror> in an answer is shown as text and never becomes an element; the test fails if skipHtml comes back.
• Definition lines inside top-level fenced or indented code are left verbatim.

Migrations

None.

Screenshots

None, text-only change.

[greptile-apps]

Automatic reviews are disabled for this repository.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud